This presentation is part of: Poster Session - Clinical Strategies; Judgment and Decison Making
INVESTIGATING DIFFERENTIAL ITEM FUNCTIONING IN PHYSICAL FUNCTIONING AND MENTAL HEALTH DOMAINS OF THE SF-36 AMONG POPULATIONS WITH THREE CHRONIC DISEASES (COPD, DEPRESSION, AND DIABETES)
Yanni F. Yu, MA, Andrew P. Yu, MA, MS, and Jeonghoon Ahn, Ph.D. University of Southern California, Pharmaceutical Economics and Policy, Los Angeles, CA
Purpose: Differential item functioning (DIF) occurs when respondents from different groups have different responses to an item given the same latent ability. The existence of DIF may compromise the validity of group comparison using the instrument. The purpose of this study is to investigate the presence and the extent of DIF among the items of physical functioning (PF) and mental health (MH) domains of the SF-36 regarding to types of chronic illness. Methods: The study population included three mutually-exclusive cohorts with COPD (n=779), depression (n=630), or diabetes (n=652), who were members of Southern California Kaiser Permanente and completed SF-36 between April 1994 and February 1995. MIMIC (multiple indicators, multiple causes) models were constructed to investigate the effects of covariates including age, gender, education, race, income, marital status, employment status, and presences of various chronic diseases on the underlying factor and the DIF effect of disease type on item responses. Results: Reliability coefficients were both high in PF (á=0.9) and MH (á=0.8) domains in all three disease groups. After controlling for the level of domain latent trait and other covariates, diabetes and COPD patients were more likely to endorse higher ability of moderate activity (PF2) and bending, kneeling or stooping (PF6) than depression patients, but less likely to rate higher capability of walking one mile (PF7). Diabetes and COPD patients tended to report lower frequency of and being downhearted (MH4), but diabetes patients tended to report higher frequency of being nervous (MH1), feeling down (MH2) and anxious (MH3). No significant DIF effects were found for other items. Conclusions: In our study population, three out of ten PF items and four out of five MH items of SF-36 showed strong DIF pertaining to disease type. The presence of DIF effect in the items implied the disease-related information beyond the domain they were in. The level of physical function derived from PF domain can be compared among different chronic diseases. However, comparison of mental health status based on MH domain across diseases needs to take DIF into account due to significant DIF effect on major items.
